2025 Rewind - My Best Articles That You Should Read
100K+ Views, Thousands of Likes and Comments
2025 was the year AI became personal. We didn’t just watch from the sidelines; we built agents, analyzed real data, and turned YouTube videos into working apps.
This year, you explored these new technologies with me, integrated them into your work, and leveled up. Our articles have been viewed over 100K times, and the feedback from you has been incredible.
Before we step into 2026, let’s revisit the 7 best articles from 2025 that you absolutely need to read.
1- NotebookLM & Gemini: I Built 2 AI Apps from YouTube Videos (Zero Code)
NotebookLM was connected to Gemini, and in this article, I turned a YouTube video into a working AI app in 2 minutes, no code, no setup.
What We Built?
Prompt Advisor
An AI app trained on Anthropic’s official Prompt Engineering course. It builds, fixes, and improves your prompts using techniques from the engineers who created Claude.
AI Teacher
A personalized learning app built from Stanford’s Machine Learning course. It teaches concepts, uses flashcards for memorization, runs interactive quizzes, and generates infographics.
The Magic
One button: “Develop app from attached Notebook.” That’s it. Your sources become the brain. Gemini becomes the interface.
Both apps. Zero code. Under 10 minutes each.
2. Personal AI Agents are The Future!
Why is everyone suddenly talking about agents? And how can you build your own?
The Easiest Way to Build Agents
CustomGPT’s (OpenAI)
Gemini Gems (Google)
Claude Projects (Anthropic)
I’ve developed over 40 agents for clients worldwide. The goal is always the same: turn hours-long tasks into minutes without losing quality.
Key Insight
We used Chain of Thought prompting, RAG, and persona prompting to build an AI Academic Research Assistant that tracks and digests research papers automatically.
And you’ll see a step-by-step explanation.
Market Proof
2 out of every 10 AI-related jobs on Upwork now mention agent-based, based on my analysis of 300+ job postings.
3. I Used to Read. Now I See with Nano Banana Pro
Nano Banana Pro isn’t just an image-generation model anymore. It’s a learning system.
What It Does?
- PDF to Infographic: Turn any book chapter into visual summaries using a modular JSON prompt
- YouTube to Visual: Convert Stanford courses or any video into infographics
- Modular Prompt System: Customize learning style (Feynman Technique, Mind Mapping, Cornell Method), aspect ratio, colors - one prompt, infinite variations
More Templates:
- Paper to Infographic
- Concept Explainer
- Comparison Chart
- Timeline
- How It Works flowcharts
One prompt, tailored exactly to how your brain learns best.
4. Data Project 6: I Analyzed 77,108 NYC Real Estate Sales
The data doesn’t lie about what works in New York City real estate, and what doesn’t.
In this article, we’ve analyzed 77,108 NYC Real estates, and let’s see what we’ve found.
Key Findings
- Antique Premium: 100+ year old buildings sell for 10% more per square foot than modern buildings.
- Size Matters (Inversely): Small buildings (1-5 units) command a 385% premium over large buildings (50+ units).
- Best Deals: Commercial garages at $607/sqft beat luxury condos. Elevator rentals get crushed at $92/sqft.
- Fastest Flip Zone: Westchester (36 days median turnover). Hold zone: Gravesend (265 days).
- Risk-Adjusted Returns: Staten Island = stability. The Bronx = chaos (but opportunity).
The Signal:
Small, old, and local wins in New York City.
The technical approach included Mann-Whitney U tests, polynomial regression, and Isolation Forest for anomaly detection.
5. The Only Prompt That Made My AI Honest
You’ve experienced this: AI agrees with you, then you spot a mistake, and suddenly it’s “You’re absolutely right!”. It’s called sycophancy. And I fixed it using a different approach.
The Zero-Sum Objectivity Scale
- Optimism (+1): Sees only potential
- Pessimism (-1): Sees only risk
- Objectivity (0): The balance we need
The Prompt
Forces AI to simulate 10 rounds of debate between an Optimist and a Pessimist until they cancel each other out.
Automation
I showed you how to build this into Claude Skills, CustomGPT, and Gemini Gems, so you never have to copy-paste again.
No more parrot AI.
6. Google’s New AI Feature Reads You Like a Book
I detected thousands of emails to delete and created a full report in seconds using Google’s new Deep Research feature.
What It Does?
Perform deep research across your entire Gmail inbox and Google Drive.
Gmail Use Case
I prompted through my Gmail and deleted the articles, using tags.
Google Drive Use Case
I prompted through my GDrive and did data analysis by just prompting.
If you’re using Gmail for business, this is a game-changer for inbox management and file organization.
7. If You Want to Learn Statistics Without Boring, Read This!
Statistics isn’t about memorizing formulas. It’s about understanding why everyday things fool your brain.
Mind-Bending Examples:
- Birthday Paradox: In a room of 30 people, there’s a 70% chance two share a birthday. Your brain thinks 10%.
- Monty Hall Problem: Switching doors doubles your chances of winning. Mathematics says, “Switch, you fool!”
- Hot Hand Fallacy: Your friend made 5 shots in a row. His next shot? Still 50-50. Streaks are illusions.
The Takeaway
Your gut feeling is wrong 70% of the time. These concepts separate AI enthusiasts from experts.
No formulas. No Greek letters. Just examples that change how you see the world.
Final Thoughts
2025 was fast. AI moved faster.
We built agents, analyzed 77K real estate transactions, turned sycophantic AI honest, and created apps from YouTube videos, all without writing code.
Our paid subscribers got these insights first, plus access to The Vault with prompts, CustomGPTs, AI tools and data project source codes.
Thanks for being a subscriber to me! Happy New Year & Christmas to you all!